This invention relates to a multi-participant, mixed-initiative voice interaction system. Existing interactive voice response (IVR) systems are designed to carry out appropriate actions in response to spoken or audible information. Examples of these systems include voice-driven phone directory services, and automated customer support systems. The interaction is driven by time and the human participant.
A single-participant voice-interaction system is effective for information retrieval activities, but not always the best choice for scenarios in which events are generated from sources or participants other than the individual providing speech input. Modeling multi-participant systems is important to clearly define and describe system behavior. Traditional IVR systems can be modeled effectively using call tree structures, with decision points based on human spoken commands. Unfortunately, using the same call-tree modeling process for multi-participant voice interaction systems results in overly complex decision points and redundancies.
Further, in some IVR systems, voice instructions from the system are cumbersome and inefficient. For example, external, variable delays often exist when retrieving information. Examples of these delays include accessing variable-length message content from a mobile device, or accessing information across a non-deterministic (wireless) network. Providing any human participants with verbal status information is important, but not always necessary. For example, the spoken “please wait” message sometimes takes longer to read out loud than the time typically required to retrieve the desired information. This is inefficient and makes the interface less pleasant for the user.